The Birth of Crowdsourcing?

From p. 106 of the first paperback edition of The Professor and the Madman, a slightly overwrought but enjoyable history of the origins of the Oxford English Dictionary, found on the shelf of a vacation rental:

The new venture that [Richard Chenevix] Trench seemed now to be proposing would demonstrate not merely the meaning but the history of meaning, the life story of each word. And that would mean the reading of everything and the quoting of everything that showed anything of the history of the words that were to be cited. The task would be gigantic, monumental, and—according to the conventional thinking of the times—impossible.

Except that here Trench presented an idea, an idea that—to those ranks of conservative and frock-coated men who sat silently in the [London Library] on that dank and foggy evening [in 1857]—was potentially dangerous and revolutionary. But it was the idea that in the end made the whole venture possible.

The undertaking of the scheme, he said, was beyond the ability of any one man. To peruse all of English literature—and to comb the London and New York newspapers and the most literate of the magazines and journals—must be instead “the combined action of many.” It would be necessary to recruit a team—moreover, a huge one—probably comprising hundreds and hundreds of unpaid amateurs, all of them working as volunteers.

The audience murmured with surprise. Such an idea, obvious though it may sound today, had never been put forward before. But then, some members said as the meeting was breaking up, it did have some real merit.

And here’s what that crowdsourcing process ended up looking like in practice:

[Frederick] Furnivall then issued a circular calling for volunteer readers. They could select from which period of history they would like to read books—from 1250 to 1526, the year of the New English Testament; from then to 1674, the year when Milton died; or from 1674 to what was then the present day. Each period, it was felt, represented the existence of different trends in the development of the language.

The volunteers’ duties were simple enough, if onerous. They would write to the society offering their services in reading certain books; they would be asked to read and make word-lists of all that they read, and would then be asked to look, super-specifically, for certain words that currently interested the dictionary team. Each volunteer would take a slip of paper, write at its top left-hand side the target word, and below, also on the left, the date of the details that followed: These were, in order, the title of the book or paper, its volume and page number, and then, below that, the full sentence that illustrated the use of the target word. It was a technique that has been undertaken by lexicographers to the present day.

Herbert Coleridge became the first editor of what was to be called A New English Dictionary on Historical Principles. He undertook as his first task what may seem prosaic in the extreme: the design of a small stack of oak-board pigeonholes, nine holes wide and six high, which could accommodate the anticipated sixty to one hundred thousand slips of paper that would come in from the volunteers. He estimated that the first volume of the dictionary would be available to the world within two years. “And were it not for the dilatoriness of many contributors,” he wrote, clearly in a tetchy mood, “I should not hesitate to name an earlier period.”

Everything about these forecasts was magnificently wrong. In the end more than six million slips of paper came in from the volunteers; and Coleridge’s dreamy estimate that it might take two years to have the first salable section of the dictionary off the presses—for it was to be sold in parts, to help keep revenues coming in—was wrong by a factor of ten. It was this kind of woefully naive underestimate—of work, of time, of money—that at first so hindered the dictionary’s advance. No one had a clue what they were up against: They were marching blindfolded through molasses.

So, even with all those innovations, this undertaking also produced a textbook example of the planning fallacy. I wonder how quickly and cheaply the task could have been completed with Mechanical Turk, or with some brush-clearing assistance from text mining?

Advertisements

Wisdom of Crowds FTW

I’m a cyclist who rides indoors a fair amount, especially in cold or wet weather. A couple of months ago, I bought an indoor cycle with a flywheel and a power meter. For the past several years, I’d been using the kind of trainer you attach to the back wheel of your bike for basement rides. Now, though, my younger son races, so I wanted something we could both use without too much fuss, and his coach wants to see power data from his home workouts.

To train properly with a power meter, I need to benchmark my current fitness. The conventional benchmark is Functional Threshold Power (FTP), which you can estimate from your average power output over a 20-minute test. To get the best estimate, you need to go as hard as you can for the full 20 minutes. To do that, you need to pace yourself. Go out too hard and you’ll blow up partway through. Go out too easy and you’ll probably end up lowballing yourself.

Once you have an estimate of your FTP, that pacing is easy to do: just ride at the wattage you expect to average. But what do you do when you’re taking the test for the first time?

I decided to solve that problem by appealing to the wisdom of the crowd. When I ride outdoors, I often ride with the same group, and many of those guys train with power meters. That means they know me and they know power data. Basically, I had my own little panel of experts.

Early this week, I emailed that group, told them how much I weigh (about 155 lbs), and asked them to send me estimates of the wattage they thought I could hold for 20 minutes. Weight matters because power covaries with it. What the other guys observe is my speed, which is a function of power relative to weight. So, to estimate power based on observed speed, they need to know my weight, too.

I got five responses that ranged from 300 to 350. Based on findings from the Good Judgment Project, I decided to use the median of those five guesses—314—as my best estimate.

I did the test on Tuesday. After 15 minutes of easy spinning, I did 3 x 30 sec at about 300W with 30 sec easy in between, then another 2 min easy, then 3 min steady above 300W, then 7 min easy, and then I hit it. Following emailed advice from Dave Guttenplan, who sometimes rides with our group, I started out a little below my target, then ramped up my effort after about 5 min. At the halfway point, I peeked at my interval data and saw that I was averaging 310W. With 5 min to go, I tried to up the pace a bit more. With 1 min to go, I tried to dial up again and found I couldn’t go much harder. No finish-line sprint for me. When the 20-minute mark finally arrived, I hit the “interval” button, dialed the resistance down, and spent the next minute or so trying not to barf—a good sign that I’d given it just about all I had.

And guess what the final average was: 314!

Now, you might be thinking I tried to hit that number because it makes for a good story. Of course I was using the number as a guideline, but I’m as competitive as the next guy, so I was actually pretty motivated to outperform the group’s expectations. Over the last few minutes of the test, I was getting a bit cross-eyed, too, and I don’t remember checking the output very often.

This result is also partly coincidence. Even the best power meters have a margin of error of about 2 percent, and that’s assuming they’re properly calibrated. So the best I can say is that my average output from that test was probably around 314W, give or take several watts.

Still, as an applied stats guy who regularly works with “wisdom of crowds” systems, I thought this was a great illustration of those methods’ utility. In this case, the remarkable accuracy of the crowd-based estimate surely had a lot to do with the crowd’s expertise. I only got five guesses, but they came from people who know a lot about me as a rider and whose experience training with power and looking at other riders’ numbers has given them a strong feel for the distribution of these stats. If I’d asked a much bigger crowd who didn’t know me or the data, I suspect the estimate would have missed badly (like this one). Instead, I got just what I needed.

Forecasting Round-Up No. 7

1. I got excited when I heard on Twitter yesterday about a machine-learning process that turns out to be very good at predicting U.S. Supreme Court decisions (blog post here, paper here). I got even more excited when I saw that the guys who built that process have also been running a play-money prediction market on the same problem for the past several years, and that the most accurate forecasters in that market have done even better than that model (here). It sounds like they are now thinking about more rigorous ways to compare and cross-pollinate the two. That’s part of what we’re trying to do with the Early Warning Project, so I hope that they do and we can learn from their findings.

2. A paper in the current issue of the Journal of Personality and Social Psychology (here, but paywalled; hat-tip to James Igoe Walsh) adds to the growing pile of evidence on the forecasting power of crowds, with an interesting additional finding on the willingness of others to trust and use those forecasts:

We introduce the select-crowd strategy, which ranks judges based on a cue to ability (e.g., the accuracy of several recent judgments) and averages the opinions of the top judges, such as the top 5. Through both simulation and an analysis of 90 archival data sets, we show that select crowds of 5 knowledgeable judges yield very accurate judgments across a wide range of possible settings—the strategy is both accurate and robust. Following this, we examine how people prefer to use information from a crowd. Previous research suggests that people are distrustful of crowds and of mechanical processes such as averaging. We show in 3 experiments that, as expected, people are drawn to experts and dislike crowd averages—but, critically, they view the select-crowd strategy favorably and are willing to use it. The select-crowd strategy is thus accurate, robust, and appealing as a mechanism for helping individuals tap collective wisdom.

3. Adam Elkus recently spotlighted two interesting papers involving agent-based modeling (ABM) and forecasting.

  • The first (here) “presents a set of guidelines, imported from the field of forecasting, that can help social simulation and, more specifically, agent-based modelling practitioners to improve the predictive performance and the robustness of their models.”
  • The second (here), from 2009 but new to me, describes an experiment in deriving an agent-based model of political conflict from event data. The results were pretty good; a model built from event data and then tweaked by a subject-matter expert was as accurate as one built entirely by hand, and the hybrid model took much less time to construct.

4. Nautilus ran a great piece on Lewis Fry Richardson, a pioneer in weather forecasting who also applied his considerable intellect to predicting violent conflict. As the story notes,

At the turn of the last century, the notion that the laws of physics could be used to predict weather was a tantalizing new idea. The general idea—model the current state of the weather, then apply the laws of physics to calculate its future state—had been described by the pioneering Norwegian meteorologist Vilhelm Bjerknes. In principle, Bjerkens held, good data could be plugged into equations that described changes in air pressure, temperature, density, humidity, and wind velocity. In practice, however, the turbulence of the atmosphere made the relationships among these variables so shifty and complicated that the relevant equations could not be solved. The mathematics required to produce even an initial description of the atmosphere over a region (what Bjerknes called the “diagnostic” step) were massively difficult.

Richardson helped solve that problem in weather forecasting by breaking the task into many more manageable parts—atmospheric cells, in this case—and thinking carefully about how those parts fit together. I wonder if we will see similar advances in forecasts of social behavior in the next 100 years. I doubt it, but the trajectory of weather prediction over the past century should remind us to remain open to the possibility.

5. Last, a bit of fun: Please help Trey Causey and me forecast the relative strength of this year’s NFL teams by voting in this pairwise wiki survey! I did this exercise last year, and the results weren’t bad, even though the crowd was pretty small and probably not especially expert. Let’s see what happens if more people participate, shall we?

A Coda to “Using GDELT to Monitor Atrocities, Take 2”

I love doing research in the Internet Age. As I’d hoped it would, my post yesterday on the latest iteration of our atrocities-monitoring system in the works has already sparked a lot of really helpful responses. Some of those responses are captured in comments on the post, but not all of them are. So, partly as a public good and partly for my own record-keeping, I thought I’d write a coda to that post enumerating the leads it generated and some of my reactions to them.

Give the Machines Another Shot at It

As a way to reduce or even eliminate the burden placed on our human(s) in the loop, several people suggested something we’ve been considering for a while: use machine-learning techniques to develop classifiers that can be used to further reduce the data left after our first round of filtering. These classifiers could consider all of the features in GDELT, not just the event and actor types we’re using in our R script now. If we’re feeling really ambitious, we could go all the way back to the source stories and use natural-language processing to look for additional discriminatory power there. This second round might not eliminate the need for human review, but it certainly could lighten the load.

The comment threads on this topic (here and here) nicely capture what I see as the promise and likely limitations of this strategy, so I won’t belabor it here. For now, I’ll just note that how well this would work is an empirical question, and it’s one we hope to get a chance to answer once we’ve accumulated enough screened data to give those classifiers a fighting chance.

Leverage GDELT’s Global Knowledge Graph

Related to the first idea, GDELT co-creator Kalev Leetaru has suggested on a couple of occasions that we think about ways to bring the recently-created GDELT Global Knowledge Graph (GKG) to bear on our filtering task. As Kalev describes in a post on the GDELT blog, GKG consists of two data streams, one that records mentions of various counts and another that captures connections  in each day’s news between “persons, organizations, locations, emotions, themes, counts, events, and sources.” That second stream in particular includes a bunch of data points that we can connect to specific event records and thus use as additional features in the kind of classifiers described under the previous header. In response to my post, Kalev sent this email to me and a few colleagues:

I ran some very very quick numbers on the human coding results Jay sent me where a human coded 922 articles covering 9 days of GDELT events and coded 26 of them as atrocities. Of course, 26 records isn’t enough to get any kind of statistical latch onto to build a training model, but the spectral response of the various GKG themes is quite informative. For events tagged as being an atrocity, themes such as ETHNICITY, RELIGION, HUMAN_RIGHTS, and a variety of functional actors like Villagers, Doctors, Prophets, Activists, show up in the top themes, whereas in the non-atrocities the roles are primarily political leaders, military personnel, authorities, etc. As just a simple example, the HUMAN_RIGHTS theme appeared in just 6% of non-atrocities, but 30% of atrocities, while Activists show up in 33% of atrocities compared with just 4% of non-atrocities, and the list goes on.

Again, 26 articles isn’t enough to build a model on, but just glancing over the breakdown of the GKG themes for the two there is a really strong and clear breakage between the two across the entire set of themes, and the breakdown fits precisely what baysean classifiers like (they are the most accurate for this kind of separation task and outperform SVM and random forest).

So, Jay, the bottom line is that if you can start recording each day the list of articles that you guys review and the ones you flag as an atrocity and give me a nice dataset over time, should be pretty easy to dramatically filter these down for you at the very least.

As I’ve said throughout this process, its not that event data can’t do what is needed, its that often you have to bring additional signals into the mix to accomplish your goals when the thing you’re after requires signals beyond what the event records are capturing.

What Kalev suggests at the end there—keep a record of all the events we review and the decisions we make on them—is what we’re doing now, and I hope we can expand on his experiment in the next several months.

Crowdsource It

Jim Walsh left a thoughtful comment suggesting that we crowdsource the human coding:

Seems to me like a lot of people might be willing to volunteer their time for this important issue–human rights activists and NGO types, area experts, professors and their students (who might even get some credit and learn about coding). If you had a large enough cadre of volunteers, could assign many (10 or more?) to each day’s data and generate some sort of average or modal response. Would need someone to organize the volunteers, and I’m not sure how this would be implemented online, but might be do-able.

As I said in my reply to him, this is an approach we’ve considered but rejected for now. We’re eager to take advantage of the wisdom of interested crowds and are already doing so in big ways on other parts of our early-warning system, but I have two major concerns about how well it would work for this particular task.

The first is the recruiting problem, and here I see a Catch-22: people are less inclined to do this if they don’t believe the system works, but it’s hard to convince them that the system works if we don’t already have a crowd involved to make it go. This recruiting problem becomes especially acute in a system with time-sensitive deliverables. If we promise daily updates, we need to produce daily updates, and it’s hard to do that reliably if we depend on self-organized labor.

My second concern is the principal-agent problem. Our goal is to make reliable and valid data in a timely way, but there are surely people out there who would bring goals to the process that might not align with ours. Imagine, for example, that Absurdistan appears in the filtered-but-not-yet-coded data to be committing atrocities, but citizens (or even paid agents) of Absurdistan don’t like that idea and so organize to vote those events out of the data set. It’s possible that our project would be too far under the radar for anyone to bother, but our ambitions are larger than that, so we don’t want to assume that will be true. If we succeed at attracting the kind of attention we hope to attract, the deeply political and often controversial nature of our subject matter would make crowdsourcing this task more vulnerable to this kind of failure.

Use Mechanical Turk

Both of the concerns I have about the downsides of crowdsourcing the human-coding stage could be addressed by Ryan Briggs’ suggestion via Twitter to have Amazon Mechanical Turk do it. A hired crowd is there when you need it and (usually) doesn’t bring political agendas to the task. It’s also relatively cheap, and you only pay for work performed.

Thanks to our collaboration with Dartmouth’s Dickey Center, the marginal cost of the human coding isn’t huge, so it’s not clear that Mechanical Turk would offer much advantage on that front. Where it could really help is in routinizing the daily updates. As I mentioned in the initial post, when you depend on human action and have just one or a few people involved, it’s hard to establish a set of routines that covers weekends and college breaks and sick days and is robust to periodic changes in personnel. Primarily for this reason, I hope we’ll be able to run an experiment with Mechanical Turk where we can compare its cost and output to what we’re paying and getting now and see if this strategy might make sense for us.

Don’t Forget About Errors of Omission

Last but not least, a longtime colleague had this to say in an email reacting to the post (hyperlinks added):

You are effectively describing a method for reducing errors of commission, events coded by GDELT as atrocities that, upon closer inspection, should not be. It seems like you also need to examine errors of omission. This is obviously harder. Two possible opportunities would be to compare to either [the PITF Worldwide Atrocities Event Data Set] or to ACLED.  There are two questions. Is GDELT “seeing” the same source info (and my guess is that it is and more, though ACLED covers more than just English sources and I’m not sure where GDELT stands on other languages). Then if so (and there are errors of omission) why aren’t they showing up (coded as different types of events or failed to trigger any coding at all)[?]

It’s true that our efforts so far have focused almost exclusively on avoiding errors of commission, with the important caveat that it’s really our automated filtering process, not GDELT, that commits most of these errors. The basic problem for us is that GDELT, or really the CAMEO scheme on which it’s based, wasn’t designed to spot atrocities per se. As a result, most of what we filter out in our human-coding second stage aren’t things that were miscoded by GDELT. Instead, they’re things that were properly coded by GDELT as various forms of violent action but upon closer inspection don’t appear to involve the additional features of atrocities as we define them.

Of course, that still leaves us with this colleague’s central concern about errors of omission, and on that he’s absolutely right. I have experimented with different actor and event-type criteria to make sure we’re not missing a lot of events of interest in GDELT, but I haven’t yet compared what we’re finding in GDELT to what related databases that use different sources are seeing. Once we accumulate a few month’s worth of data, I think this is something we’re really going to need to do.

Stay tuned for Take 3…

What Should the U.S. Do Now in Syria?

You tell me.

To help you do that, I’ve created a pairwise wiki survey on All Our Ideas. Click HERE to participate. You can vote on the options I listed or add your own.

Results are updated in real time. Just click on the View Results tab to see what the crowd is saying so far.

Before you add an idea, make sure it isn’t already covered in the existing set by clicking on the View Results tab and then the View All button at the bottom of the list.

  • Author

  • Follow me on Twitter

  • Follow Dart-Throwing Chimp on WordPress.com
  • Enter your email address to follow this blog and receive notifications of new posts by email.

    Join 13,629 other followers

  • Archives

  • Advertisements
%d bloggers like this: